Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Quackerjack v3

A Comment Thread Analyzer and Report Generator

A CLI script and web service, written in Go, that takes a YouTube video URL and generates a report about the content of the comment stream.


The report includes:

  • The total number of comments and how many were collected.
  • The top keywords from the comments.
  • The sentiment analysis of the comments.
  • Sentiment over the first 30 days from upload.
  • Emoji usage analysis.

v3 Changes:

  • New interface!
  • Mobile friendly!
  • No more redis!
  • Emoji analysis!

NOTE: Submissions of training data is more than welcome! YouTube comments are a hard thing to analyze so the more data we can collect the better! Please submit a pull request or you can reach out to me on Twitter: @thatmikeflynn


go get


The script can be run via the command line, which will return a JSON object of the resulting data:

quackerjack -stopwords ./static/stopwords.txt[,/comma/delimited/textfiles] -post [post url] [-verbose]

You can also run quackerjack as a web service. Add the -server flag to start a web server that has two endpoints on port 8080: / is a simple web interface where you can enter post urls and it displays the results in a visual report. /api is the JSON web service route which takes a ?vid= param and retuns a JSON response.


-post XXXX YouTube, or Instagram post URL.

-ytkey xxxxxxxxxxxxxxxx Your Google API key with access to the YouTube Data API.

-stopwords /xxx/yy/zz.txt A comma delimitated list of additional stop word text files not in config.

-verbose Standard boolean flag for extra logging to std out.

-server Start the web interface rather than having the JSON get dumpped to std out.

-port 8000 Override the web server port.


The new internal analysis engine is trained automatically when the Quackerjack service is started. The training documents are located in static/training and are simply a line of text and a 1 (positive) or 0 (negative) separated by a pipe. To improve the analysis, simply add as many training lines as you'd like and restart Quackerjack (from the git repo directory).

app_1  | 2020/04/28 15:56:08 Loading conf file: /go/src/quackerjack/quackerjack-docker.conf
app_1  | 2020/04/28 15:56:08 Num of docs in TRAIN dataset: 2730
app_1  | 2020/04/28 15:56:08 Num of docs in TEST dataset: 310
app_1  | 2020/04/28 15:56:09 SentimentClassifier.totalWords: 17925
app_1  | 2020/04/28 15:56:09 Accuracy on TEST dataset is 83.5 with 1.9 unknowns
app_1  | 2020/04/28 15:56:09
app_1  | Accuracy on TRAIN dataset is 94.0 with 0.0 unknowns
app_1  | 2020/04/28 15:56:09 Web server running on 8000


  • go 1.14
  • Docker

If you want to dig in to the code, you can clone this repo...

git clone

I've included a Docker image and docker-compose config that will run the web service in verbose mode by default. You can also run make docker-shell to get a shell on the Docker instance for active development.

Don't for get to update the saved assets if you update the web interface: make data !

To Do:

  • Save a copy of the resulting model for easy go get download and startup.

Who is Quackerjack?

He's a villain from Darkwing Duck.

You can’t perform that action at this time.